[C++][Parquet] Implement specification-compliant floating point comparison

[asfimport]

Currently, the floating point comparison order defined by parquet-format is ambigous and different implementations use different orderings. Once the specification is made unambigous (PARQUET-1222), implementations should be updated to conform to it.

Reporter: Zoltan Ivanfi / @zivanfi

Related issues:

• Specify a well-defined sorting order for float and double types (is blocked by)

_{Note: This issue was originally created as PARQUET-1224. Please see the migration documentation for further details.}

[asfimport]

Gang Wu / @wgtmac:
This is fixed in [PARQUET-1766: [C++] Handle parquet::Statistics NaNs and -0.0f as per upstream parquet-mr by fsaintjacques · Pull Request #6167 · apache/arrow (github.com)|https://github.com//pull/6167]